The Atlan MCP Cookbook

Document SQL with business context

Generate business-readable descriptions for SQL models from the context already in Atlan.

All recipes

On the menu

Use case
Make data understandable
Action
Make changes
Persona
AI Builder, Data Engineer
Effort
Moderate
Runs in
any AI tool

The task

Scenario. Your monthly_revenue_summary dbt model is documented in the way models usually are: a note about the SQL, nothing about what it means to the business. You want a description a stakeholder can read, built from the glossary terms and descriptions its source tables already carry in Atlan.

The ask

  • The Atlan descriptions and glossary terms for the model's source tables
  • A business-friendly description of what the model calculates and who it is for
  • The description saved back to the model in Atlan

Hero prompt

Hero prompt

For the monthly_revenue_summary model in dbt, look up the Atlan descriptions and glossary terms for its source tables and write a business-friendly description explaining what it calculates and who it's for.

Step by step

  1. Retrieve glossary terms and descriptions for the source assets.

    Step 1

    Look up the Atlan descriptions and glossary terms for <model>'s source tables.

  2. Generate a business-readable description for the model.

    Step 2

    Write a business-friendly description of what <model> calculates and who it's for.

  3. Write the description back to the asset in Atlan.

    Step 3

    Save it to the <model> asset in Atlan.

What you'll give it

Required context

  • The model to document (monthly_revenue_summary)
  • Glossary terms that exist for its source assets
  • Permission to write the description back

Optional context

  • The audience for the description (an exec reader versus an analyst)
  • A house format, such as description plus "used by" plus grain

What Atlan creates

Outputs

  • The descriptions and glossary terms of the model's source tables, pulled together
  • A business-readable description of what the model calculates and who uses it
  • The description saved to the model in Atlan

Summary. The model gets documentation a stakeholder can actually read, built from context you already curated, so the description writes itself once the glossary is in place.

Writes to Atlan

Changed: the description on the monthly_revenue_summary model. Left untouched: the source tables and their terms, which are read to write the description, and the model's SQL.

Follow-up prompts

Keep going once the first answer lands.

Do the layer

Do the layer

Do the same for every model in the revenue folder that has no business description.

Add the grain

Add the grain

Extend the description with the model's grain and refresh cadence.

Flag the gaps

Flag the gaps

If a source table has no glossary term, tell me, so I can define it before documenting the model.

Tips & troubleshooting

Cook's note

Start from the source tables' terms. The description writes itself once the glossary is there.

  • Match the tool to the work. Writing readable prose is close-reading work; use a more capable model, review the draft, then save.
  • Automate or chain it. Chain it to your dbt build: on a new or changed model, draft the business description for review before it ships.
  • Ask what it changed. Add "confirm which model you wrote the description to." A one-line check that the write landed on the right asset.
THE ATLAN MCP COOKBOOK

Point your AI at the Atlan MCP.
Run your first recipe.

[Website env: production]